Digital Twin for Li-ion Batteries

Table of Contents

1. Project Overview
2. Architecture
3. Components
4. Setup Instructions
5. Features
6. Future Enhancements

Project Overview

This project implements a digital twin for a Lithium-ion battery system using AWS cloud services. It simulates IoT device data, processes it in real-time, applies machine learning for predictive analytics, and visualizes the results. The system demonstrates proficiency in cloud architecture, real-time data processing, IoT systems, and machine learning integration.

Key features:

• Real-time IoT data simulation and ingestion
• Scalable data processing pipeline
• Machine learning integration for battery life prediction
• Interactive data visualizations
• Infrastructure as Code (IaC) using Terraform
• Monitoring and logging with Amazon CloudWatch

Architecture

The project utilizes a serverless architecture on AWS, leveraging the following services:

• AWS IoT Core
• Terraform for IaC
• Amazon CloudWatch
• Amazon QuickSight
• Amazon SageMaker
• Amazon Redshift
• Amazon S3
• Amazon EC2

code used to create this architecture diagram

Components

Component	Purpose	Data Sources	Data Sent
IoT Devices / Data Simulator	Simulate real-time battery data	Battery sensors/ scripts to simulate sensor data	Raw battery data (voltage, current, temperature, etc.) via MQTT
AWS IoT Core	Manage IoT device connections and data ingestion	IoT Devices/ scripts simulating IoT devices	Raw battery data to EC2
EC2 - Data Processor	Process and transform raw data	AWS IoT Core	Processed data to Redshift
Amazon Redshift	Store processed data	EC2	Processed data to QuickSight
Amazon SageMaker	Train and host ML models	S3 (training data)	ML model predictions to EC2
Amazon QuickSight	Create and host data visualizations	Redshift	Visualizations to users
Terraform	Manage infrastructure as code	N/A	Resource definitions to AWS
Amazon CloudWatch	Monitor and log system performance	All AWS services	Alerts, logs to administrators

Setup Instructions

Create a notebook on Amazon SageMaker

Open Jupyter Notebook

Train the model with historical data stored in S3 bucket

in case you want to use the model directly within the EC2 instance, you can move the model.tar.gz file to your EC2 instance from the corresponding output folder of the sagemaker instance and unzip it to obtain the model.pth file

Make sure your EC2 instance has all these files

Subscribe to battery/data topic on MQTT test client

Run the batterydatasimulator.py and the batterydatapredictor.py simulatenously

Verify the data being loaded in the Redshift table

Connect the Redshift table to Quicksight to generate Real Time Visualization for Analysis and Monitoring

Features

Data Processing and Machine Learning

• Real-time data ingestion from simulated IoT devices
• Machine learning model trained on SageMaker for battery life prediction
• Data processing on EC2 instance

Visualizations

QuickSight dashboards provide the following visualizations:

1. Real-time battery voltage, current, and temperature
2. Battery capacity over time and cycle number
3. Predicted vs actual capacity degradation
4. Temperature distribution across battery cycles

Deployment

The project uses Terraform for infrastructure as code:

1. Infrastructure defined in Terraform configuration files
2. Terraform provisions and manages AWS resources
3. Changes to infrastructure managed through version control

Monitoring and Maintenance

• CloudWatch is used for monitoring all AWS services
• CloudWatch Alarms are set up for critical metrics (e.g., battery temperature exceeding thresholds)
• CloudWatch Logs collect logs from all services for troubleshooting

Future Enhancements

1. Implement anomaly detection using Amazon SageMaker
2. Add support for multiple battery types
3. Implement a web-based dashboard for real-time monitoring
4. Integrate with a physical battery testing setup for real-world data
5. Implement containerization using Docker and container orchestration with Kubernetes
6. Develop a CI/CD pipeline for automated testing and deployment
7. Explore cloud-agnostic solutions for portability across different cloud providers